A proposal of a Method for Extracting Strategy Phrases Considering with Dependency Analysis from IR Documents of Enterprises

Abstract

It have been attempted to mechanically extract the strategy-words/phrases which are valuable for analysis of enterprise management from IR documents published by the enterprise. In preceding studies, the method which could extract coined words or unique strategy-words/phrases by applying stratified N-Gram analysis with words frequency, Gram length, and inclusions of N+1 Gram words, has been proposed. However, the extracted results by the method might include some valueless words/phrases as strategy-words. In this study, we address that the considerations of the sentence structure as well as part-of-speeches is effective to improve accuracy, after analyzing the incorrectly extracted words/phrases. Furthermore, we propose a new method to extract strategy-phrases with higher accuracy by applying the result of dependency analysis.